Focused on the issue that the traditional static detection and dynamic detection methods cannot cope with malicious PDF document attacks using a lot of obfuscation and unknown technologies， a new detection model based on system calls and data provenance， called NtProvenancer， was proposed. Firstly， the system call records during execution of the document were collected by the system call tracing tool. Then， the data provenance technology was used to establish a data provenance graph based on the system calls. After that， the feature segments of system calls were extracted for detection by using the key point algorithm of the graph. The experimental dataset consists of 528 benign PDF documents and 320 malicious ones. The test was carried out on Adobe Reader， and the Term Frequency-Inverse Document Frequency （TF-IDF） and the rarity algorithm in PROVDETECTOR were used to replace the key point algorithm of the graph to conduct the comparative study. The results show that NtProvenancer has better performance on precision and F1 score. Under the optimal parameter setting， the proposed model has the average time of document training and detection stages of 251.51 ms and 60.55 ms respectively， the false alarm rate lower than 5.22%， and the F1 score reached 0.989， showing that NtProvenancer is an efficient and practical model for PDF document detection.

Aiming at that Long Short Term Memory (LSTM) has only one pedestrian considered in isolation and cannot realize prediction with various possibilities, an attention mechanism based generative model for pedestrian trajectory prediction called AttenGAN was proposed to construct pedestrian interaction model and predict multiple reasonable possibilities. The proposed model was composed of a generator and a discriminator. The generator predicted multiple possible future trajectories according to pedestrian's past trajectory probability while the discriminator determined whether the trajectories were really existed or generated by the discriminator and gave feedback to the generator, making predicted trajectories obtained conform social norm more. The generator consisted of an encoder and a decoder. With other pedestrians information obtained by the attention mechanism as input, the encoder encoded the trajectories of the pedestrian as an implicit state. Combined with Gaussian noise, the implicit state of LSTM in the encoder was used to initialize the implicit state of LSTM in the decoder and the decoder decoded it into future trajectory prediction. The experiments on ETH and UCY datasets show that AttenGAN can provide multiple reasonable trajectory predictions and can predict the trajectory with higher accuracy compared with Linear, LSTM, S-LSTM (Social LSTM) and S-GAN (Social Generative Adversarial Network) models, especially in scenes of dense pedestrian interaction. Visualization of predicted trajectories obtained by the generator indicated the ability of this model to capture the interaction pattern of pedestrians and jointly predict multiple reasonable possibilities.

According to the gray distribution characteristics of microaneurysms, a new microaneurysm detection algorithm was proposed. First, by multi-scale matched filtering, candidate microaneurysm lesions were picked out as seed points. And region growing technology was applied to segment the lesion areas. Then the features of the lesion areas were extracted. Finally the Adaboost neural network ensemble was designed to distinguish the real microaneurysm from all of the candidate lesions. The proposed method was tested on public ROC database. The experimental results show that the average detection accuracy is 40.92%, which is better than that of previous doublering filtering and morphological methods.

An enhanced algorithm associating AprioriTid with transaction reduction and item reduction technique was put forward. In the algorithm candidate set generation and the support calculation of each itemset were created after each transaction was compressed and connected, and the key word identifying was adopted in the candidate set, thus the process of pruning and string pattern matching was removed from AprioriTid algorithm. Testing results showed that the algorithm clearly outperformed AprioriTid algorithm.